Tom Tango explains the various pitcher WARs found at FanGraphs, and makes this salient observation:

For me: I just take a position half-way between fWAR and rWAR, when I need something quick.

When I worked in information retrieval, one could improve search results by taking the rankings of various search engines, and combining those rankings through, for example, a Borda count. Since WAR is a framework, and how you calculate various run values can change from implementation to implementation, the average WAR of the implementations is probably better than any single WAR value. I’d actually like to see more WAR implemenations, so we get a better combined result.